Security audit
Security checks for vulnerabilities and agentic risk
The skill is an agricultural media-analysis wrapper, but it also performs under-disclosed cloud identity creation, token storage, and remote history access tied to user media.
Install only if you are comfortable sending farm imagery or video, URLs, and identity-linked report queries to the lifeemergence.com services, and with the skill creating/reusing a workspace user database that may store API tokens. Review the publisher and service terms before using it with sensitive farm, location, or business data.
Authenticated API access, identity creation/persistence, local file I/O, and environment detection materially affect privacy and security posture but are not reflected in the agricultural-analysis description. In context, this makes the skill more dangerous because users may upload potentially sensitive farm imagery without realizing it is tied to internal identities and handled by remote services.
Authenticated API access, identity creation/persistence, local file I/O, and environment detection materially affect privacy and security posture but are not reflected in the agricultural-analysis description. In context, this makes the skill more dangerous because users may upload potentially sensitive farm imagery without realizing it is tied to internal identities and handled by remote services.
Authenticated API access, identity creation/persistence, local file I/O, and environment detection materially affect privacy and security posture but are not reflected in the agricultural-analysis description. In context, this makes the skill more dangerous because users may upload potentially sensitive farm imagery without realizing it is tied to internal identities and handled by remote services.
Authenticated API access, identity creation/persistence, local file I/O, and environment detection materially affect privacy and security posture but are not reflected in the agricultural-analysis description. In context, this makes the skill more dangerous because users may upload potentially sensitive farm imagery without realizing it is tied to internal identities and handled by remote services.
Authenticated API access, identity creation/persistence, local file I/O, and environment detection materially affect privacy and security posture but are not reflected in the agricultural-analysis description. In context, this makes the skill more dangerous because users may upload potentially sensitive farm imagery without realizing it is tied to internal identities and handled by remote services.
Authenticated API access, identity creation/persistence, local file I/O, and environment detection materially affect privacy and security posture but are not reflected in the agricultural-analysis description. In context, this makes the skill more dangerous because users may upload potentially sensitive farm imagery without realizing it is tied to internal identities and handled by remote services.
Authenticated API access, identity creation/persistence, local file I/O, and environment detection materially affect privacy and security posture but are not reflected in the agricultural-analysis description. In context, this makes the skill more dangerous because users may upload potentially sensitive farm imagery without realizing it is tied to internal identities and handled by remote services.
Authenticated API access, identity creation/persistence, local file I/O, and environment detection materially affect privacy and security posture but are not reflected in the agricultural-analysis description. In context, this makes the skill more dangerous because users may upload potentially sensitive farm imagery without realizing it is tied to internal identities and handled by remote services.
Authenticated API access, identity creation/persistence, local file I/O, and environment detection materially affect privacy and security posture but are not reflected in the agricultural-analysis description. In context, this makes the skill more dangerous because users may upload potentially sensitive farm imagery without realizing it is tied to internal identities and handled by remote services.
Authenticated API access, identity creation/persistence, local file I/O, and environment detection materially affect privacy and security posture but are not reflected in the agricultural-analysis description. In context, this makes the skill more dangerous because users may upload potentially sensitive farm imagery without realizing it is tied to internal identities and handled by remote services.
Authenticated API access, identity creation/persistence, local file I/O, and environment detection materially affect privacy and security posture but are not reflected in the agricultural-analysis description. In context, this makes the skill more dangerous because users may upload potentially sensitive farm imagery without realizing it is tied to internal identities and handled by remote services.
Authenticated API access, identity creation/persistence, local file I/O, and environment detection materially affect privacy and security posture but are not reflected in the agricultural-analysis description. In context, this makes the skill more dangerous because users may upload potentially sensitive farm imagery without realizing it is tied to internal identities and handled by remote services.
YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).
---
name: "smyx-uav-farm-health-index-map-analysis"
description: "Using multispectral or high-resolution RGB cameras mounted on agricultural UAVs to capture orthophotos or mosaics of farmland, AI models compute vegetation indices (e.g., NDVI, NDRE) and generate a farm health-index heatmap, where colors distinguish crop vigor (red = poor, yellow = medium, green = healthy). | 通过农业无人机平台搭载的多光谱或高分辨率RGB相机，采集农田的正射影像或拼接图，利用AI模型计算植被指数（如归一化植被指数NDVI、归一化红边指数NDRE等），生成农田健康指数热力图，用颜色区分作物长�
The manifest describes analysis of multispectral or RGB orthophotos/mosaics to compute vegetation indices and generate health heatmaps. However, the code explicitly requires a local or network video path and sends it as videoUrl, indicating a broader or different media-analysis workflow than the declared image/map-analysis purpose.
The implementation materially deviates from the manifest: it performs generic video analysis and history listing rather than UAV farmland orthophoto or vegetation-index processing. This is dangerous because users and higher-level systems may trust the declared agricultural purpose while actually sending arbitrary media or identifiers to a different backend capability, creating scope mismatch, data handling surprises, and possible unauthorized collection or exfiltration of unrelated content.
The User model stores identity fields plus token and open_token values for a skill whose declared purpose is farm-image health analysis. Unnecessary credential/token storage expands the attack surface and can lead to sensitive secret exposure if the local SQLite file is accessed, copied, or reused by unrelated components.
This utility file contains identity resolution, local user persistence, token handling, workspace inspection, and authenticated outbound API logic that are unrelated to UAV farm health-map analysis. The mismatch between declared skill purpose and actual capabilities greatly expands the attack surface and enables collection or use of user identity and credentials beyond what a farm imaging skill needs.
The code can silently call an external /sys/phoneLogin endpoint with register=1 and user identifiers, effectively auto-registering or logging in users without any obvious relationship to UAV crop-health analysis. This creates undisclosed account creation and identity transmission risk, and could bind users to external services or leak identifiers without informed consent.
The skill declares broad operational behavior involving shell execution, local file handling, network access, and environment usage, but does not scope or constrain those capabilities with explicit permissions metadata. In an agent ecosystem, missing tool-scope declarations increase the chance of unintended or excessive tool use, making review and policy enforcement harder.
The default trigger is broad enough to activate the skill for generic file-analysis requests, increasing the risk of unintended execution and unnecessary transmission of user files to backend services. In an environment with networked skills and automatic attachment handling, overbroad triggers can cause privacy-impacting misfires.
A long keyword list without scope limits can cause the skill to match loosely related conversations and automatically handle files or query historical reports unexpectedly. Because this skill also performs cloud/API operations and identity-linked history lookup, accidental triggering has elevated privacy and data-minimization consequences.
The skill omits a clear warning that uploaded imagery and report queries are sent to cloud/API services, which weakens informed user consent and obscures data-flow risks. This is especially important here because the skill also links report retrieval to internal identity handling and remote history access.
The method returns a fixed Chinese string for output content, with no indication that the user can choose or opt into this locale. Per the policy, forcing a specific language without user choice is a natural-language policy violation.
The manifest describes computing vegetation indices like NDVI/NDRE from agricultural UAV imagery, but the entrypoint accepts a pet_type argument with values cat, dog, and other, and mutates ConstantEnum.DEFAULT__PET_TYPE. This indicates the implementation is repurposed from a pet/media analysis workflow rather than a dedicated farmland health-index analyzer, creating a semantic mismatch between the claimed domain-specific functionality and the actual code interface.
The CLI help text says the tool generates farmland health-index maps from UAV imagery, yet the core implementation simply calls skill.get_output_analysis(input_path) and ignores api_url, api_key, output_level, and the documented NDVI/NDRE-specific processing steps. The inline/user-facing documentation therefore overstates or mischaracterizes what this file itself does compared with the actual behavior shown here.
Detected: suspicious.install_untrusted_source